Interval Forecast based on Markov Chain Monte Carlo with Autoregressive Recurrent Networks for Natural Gas Consumption

Natural gas plays an important role in both modern life and national carbon neutral policies. In the case of companies, for example, being able to generate probabilistic forecast with confidence intervals is critical for procurement, storage and supply of natural gas. In this paper, a novel forecast method based on an autoregression recurrent network and Markov Chain Monte Carlo algorithm is proposed for natural gas consumption prediction. The application of Markov Chain Monte Carlo algorithm contributes to the implementation of deep learning techniques in decision making. The forecast accuracy of the proposed method is improved by about 5.87% compared to the state-of-the-art method through extensive empirical evaluation on the real gas dataset.